Nscale IPO Puts Its Microsoft and Anthropic Dependence on Trial

Nscale filed for an initial public offering after assembling more than $103 billion in contracts, but roughly 85% comes from Microsoft and Anthropic. The Nscale IPO therefore offers investors an unusually concentrated wager on two companies maintaining enormous demand for AI computing capacity.

The numbers behind that wager are striking. Nscale generated $140.6 million in revenue during the first half of 2026 while recording a $1.02 billion net loss. Its contracted business is much larger than its present operations, and converting that backlog requires data centers, chips, power, financing, and reliable execution.

That gap makes Nscale more than another fast-growing infrastructure company. CoreWeave reached public markets with similar questions about customer concentration, capital requirements, and dependence on a tightly connected group of AI companies. Nscale now asks Wall Street to accept those risks again, this time with even more of the investment case resting on contracts that extend years into the future.

The Nscale IPO Starts With a Huge Contract Book

Nscale is bringing public investors a contract story whose scale far exceeds the company’s current revenue base.

The London-based company filed its registration statement with the U.S. Securities and Exchange Commission on September 18, 2026. It applied to list ordinary shares on the New York Stock Exchange under the ticker NSCL.

Nscale had not specified the number of shares or an expected offering range when it announced the filing. Goldman Sachs, J.P. Morgan, and Morgan Stanley are serving as lead bookrunners, according to the company’s IPO announcement.

The filing followed a rapid transformation. Nscale emerged from Australian cryptocurrency miner Arkon Energy in 2024 and expanded into AI computing infrastructure. It now develops data centers and rents access to GPU clusters used for training and running AI models.

Companies in this segment are often called neoclouds. A neocloud is a specialized cloud provider built around high-performance GPU computing rather than a broad catalog of conventional enterprise services.

Nscale says its offering combines computing hardware, networking, storage, software, data centers, and access to power. That vertically integrated model gives the company more control over deployment, but it also increases the capital required before customers can use the capacity.

As of August 31, Nscale reported more than $103 billion in active and contracted contract value. Microsoft agreements signed between September 2025 and April 2026 provide for payments of up to approximately $43.8 billion through December 2033.

Another four agreements with Anthropic provide for aggregate payments of up to approximately $44.6 billion. Together, those two relationships represent about 85% of the contract total highlighted in the company’s filing.

Those headline amounts are not equivalent to revenue already earned. Nscale must build the required infrastructure, meet delivery conditions, maintain service availability, and keep customers committed throughout each agreement.

The distinction matters because the company remains much smaller on an accounting basis. Revenue increased from $10.4 million in the first half of 2025 to $140.6 million one year later. Its net loss expanded from approximately $369 million to $1.02 billion over the same comparison period.

That growth rate demonstrates that Nscale is activating real infrastructure. It does not establish that the company can convert tens of billions in contracted value at the expected pace or margin.

The company’s largest customer supplied 52% of revenue during the six months ending June 30, 2026. One customer had provided 73% of revenue in 2025 and substantially all revenue in 2024.

Nscale does not identify every historical customer within those concentration disclosures. However, its SEC filing explicitly warns that Microsoft and Anthropic are expected to become significant customers in future periods.

The Nscale IPO is therefore built around a sharp contrast. The company already possesses demand commitments that many infrastructure providers would struggle to match. Yet its present financial results show how early it remains in the expensive process of fulfilling them.

Microsoft and Anthropic Create Both Scale and Fragility

The same two customers that make Nscale’s backlog exceptional also create its clearest point of failure.

Customer concentration is not automatically a weak business model. Large data center projects need anchor tenants because lenders and developers require predictable demand before committing billions to construction and hardware.

Microsoft and Anthropic can provide that demand. Both need large quantities of advanced computing capacity, and both have incentives to diversify beyond infrastructure they own directly.

For Nscale, large contracts can make individual campuses financeable. Predictable customer payments can support borrowing, equipment purchases, construction schedules, and long-term power arrangements.

The problem begins when one or two counterparties carry most of the economic case. A delayed deployment, revised model strategy, financing dispute, or lower infrastructure budget can affect several Nscale projects at once.

Microsoft is especially important because its agreements with Nscale span multiple locations and extend through 2033. The size of the commitment gives Nscale a credible route to expansion, assuming the company satisfies delivery and availability requirements.

However, Microsoft is not a passive buyer. It operates Azure, develops its own data centers, works with OpenAI, and purchases capacity from other infrastructure providers. It can adjust where workloads run as economics and technical requirements change.

Anthropic introduces a different version of the risk. Its agreements support dedicated infrastructure at Nscale’s Monarch Compute Campus in West Virginia. The project is expected to use Nvidia’s next-generation Vera Rubin systems.

Nscale must first obtain qualifying financing for the GPU equipment and data center infrastructure required by those agreements. At the filing date, it had not obtained binding commitments for that financing.

The filing says Nscale is developing a plan involving multiple funding sources. It also warns that financing might not arrive on acceptable terms, or at all.

That condition turns the Anthropic contract into both an asset and an obligation. The agreement signals substantial demand, but it requires Nscale to raise money and deliver infrastructure before much of that value can become revenue.

Anthropic also retains contractual remedies if Nscale misses specified milestones or fails to maintain required service levels. Those provisions reportedly include cancellation or termination rights tied to delays and sustained underperformance.

That is normal protection for a buyer committing to essential infrastructure. It remains significant because Anthropic represents such a large portion of Nscale’s contracted value.

The concentration also links Nscale’s prospects to Anthropic’s own spending decisions. Anthropic needs continuing customer growth and access to capital to support its expanding compute commitments.

A strong Anthropic business would not eliminate this risk. It would simply make the commitment easier to defend. A strategic shift toward different chips, locations, or infrastructure partners could still reduce Nscale’s expected workload.

Microsoft and Anthropic also operate inside a densely connected financing system. Cloud providers invest in AI developers, chip companies invest in infrastructure operators, and those operators use the proceeds to buy chips.

Nvidia illustrates that relationship. It supplies the GPUs that underpin Nscale’s service and is also an investor in the company.

In September, Nvidia agreed to provide $1 billion through convertible notes or non-voting shares within a broader $3.1 billion financing arrangement. The capital can support Nscale’s expansion, while that expansion creates another buyer for Nvidia systems.

An AI financing analysis described the filing as a map of the sector’s circular financial relationships. That description captures the concern without proving that the model is unsustainable.

Strategic alignment can lower execution risk because suppliers, customers, and investors all benefit when projects launch. It can also transmit problems quickly when one participant reduces spending or misses a commitment.

For public investors, the question is not whether Microsoft and Anthropic are desirable customers. It is whether their scale compensates for the limited room Nscale has to absorb a major change in either relationship.

The Contract Value Is Not the Same as Earned Revenue

The central tradeoff in the Nscale IPO is visibility against execution: contracts clarify demand, but they do not remove construction and financing risk.

A $103 billion contract book creates an immediate impression of certainty. Yet infrastructure agreements frequently depend on delivery dates, available capacity, uptime, and other performance conditions.

Nscale can recognize revenue only as it provides contracted services. It cannot treat the total potential value of agreements running through the next decade as current sales.

This difference explains why the company’s $140.6 million of first-half revenue can coexist with contracts exceeding $103 billion. Most of the promised capacity has not entered service or has not reached full deployment.

Nscale must secure suitable sites, power, networking equipment, cooling systems, and GPUs. It must then integrate those components and operate them at the reliability levels required by sophisticated AI customers.

Each stage introduces a potential bottleneck. Electrical interconnections can take years. Construction costs can change. Advanced chips can arrive late, while newer hardware can alter a customer’s preferred configuration.

The Anthropic agreements make that dependency explicit. Nscale must use its best efforts to secure qualifying financing within a specified period, subject to any extension agreed by both companies.

Financing is not a minor administrative step. AI campuses require heavy spending before they generate recurring service revenue, leaving operators dependent on debt, customer prepayments, private capital, or public markets.

Nscale has already demonstrated access to several forms of funding. It announced a $2 billion Series C round in March 2026, led by Aker and 8090 Industries.

The company also arranged debt backed by GPUs and individual infrastructure projects. In July, an Nscale subsidiary entered facilities totaling $790 million for infrastructure spending in Norway.

These transactions show that lenders and strategic investors are willing to finance the company. They do not guarantee that every project can obtain enough capital on terms that preserve attractive returns for shareholders.

Debt raises another issue. Borrowing can accelerate construction when contracted revenue is dependable. It can also magnify losses if a campus launches late, runs below capacity, or requires refinancing under less favorable conditions.

Nscale’s first-half loss reflects the burden of scaling before its contract base becomes operating revenue. Some spending supports assets that can produce income for years, but investors must still judge the resulting cash requirements.

The company also faces technology timing risk. A facility designed around one GPU generation can encounter delays while customers prepare for the next generation.

Nscale plans to deploy Nvidia’s Vera Rubin platform for parts of its future capacity. That roadmap can make the company attractive to customers seeking newer systems. It can also tie deployment schedules to hardware that is not yet operating at full commercial scale.

Energy availability creates another layer of uncertainty. AI clusters require large, consistent power supplies, and the surrounding transmission infrastructure cannot expand as quickly as software demand.

Nscale operates or develops projects in Norway, Portugal, Texas, and West Virginia. Geographic diversity reduces dependence on a single power market, but it complicates construction, regulation, procurement, and operations.

The company’s vertical integration is intended to manage those dependencies. Owning more of the delivery chain can improve coordination across power, buildings, hardware, and cloud software.

The same strategy places more execution responsibility inside Nscale. It cannot easily blame an outside provider if delays prevent the company from meeting customer milestones.

This is why contracted value should be viewed as conditional visibility rather than guaranteed income. The contracts identify buyers, workloads, and potential payments. They do not finance or construct the required capacity by themselves.

Investors should focus on the bridge between those two states. The decisive indicators will include financed capacity, energized campuses, installed GPUs, customer acceptance, service availability, and recognized revenue.

CoreWeave Shows Why Wall Street Will Look Past Growth

CoreWeave gives investors a public-market reference for valuing fast growth alongside customer concentration, debt, and relentless capital spending.

Like Nscale, CoreWeave evolved from cryptocurrency mining into specialized AI infrastructure. It used access to Nvidia GPUs and large customer contracts to expand much faster than traditional data center operators.

CoreWeave also entered public markets with substantial dependence on Microsoft. The company disclosed that Microsoft generated approximately 67% of its 2025 revenue, according to its filings.

That concentration did not prevent CoreWeave from growing. It did ensure that investors closely tracked changes in customer mix, capital expenditure, debt, and the conversion of backlog into revenue.

Nscale’s concentration is measured differently. Its filing highlights current revenue dependence while Microsoft and Anthropic dominate future contracted value. That makes a direct percentage comparison imperfect.

However, both companies expose investors to the same fundamental mechanism. A small number of large customers helps infrastructure providers build quickly, then becomes a risk once public shareholders demand predictable results.

CoreWeave has worked to diversify through commitments from OpenAI and other customers. Its quarterly filing still identifies customer concentration, capital needs, and supplier dependence as material risks.

Nscale’s other competitors include Nebius, Lambda, and Crusoe. Each offers a different balance between owned infrastructure, leased facilities, cloud software, customer commitments, and external financing.

Crusoe recently raised $3.9 billion at a reported $30.9 billion valuation. Its projects include large data centers as well as smaller modular facilities designed for faster deployment.

Nebius provides AI cloud services while expanding infrastructure across several regions. Lambda has built its position around access to Nvidia hardware and services for AI developers.

These companies compete for more than customers. They compete for GPUs, power, land, construction capacity, employees, and affordable financing.

They also face the hyperscalers. Microsoft, Amazon, Google, and Oracle can build infrastructure internally while purchasing external capacity when demand exceeds their immediate supply.

That relationship makes neoclouds both partners and potential substitutes. A hyperscaler might rely on a specialist during a capacity shortage, then shift workloads once its own facilities become available.

Specialized providers argue that their focused architecture and deployment speed justify a durable role. Customers might also prefer multiple vendors to reduce dependence on any single cloud platform.

The public market must decide how much of today’s outsourced demand is structural. If neoclouds remain essential partners, Nscale’s long contracts offer a meaningful advantage.

If customers gradually internalize more capacity, the contracts provide only temporary protection. Renewal terms and optional extensions would then matter more than the initial headline value.

The comparison with Applied Digital adds another warning. That data center developer has also relied on a small number of large tenants, including CoreWeave and Oracle.

Long leases can make infrastructure projects bankable. They can simultaneously concentrate credit exposure and leave future growth dependent on a narrow group of counterparties.

Nscale’s advantage is the scale and prominence of its customers. Its disadvantage is that public investors have already learned to question whether concentrated AI demand deserves a premium.

Wall Street will not evaluate the Nscale IPO solely through revenue growth. Investors will test contract quality, financing progress, project economics, and the company’s ability to diversify before current agreements expire.

The result will affect its competitors as well. A favorable reception could establish another public valuation benchmark for AI infrastructure operators.

A weak reception would make private funding and future listings harder across the sector. It would signal that large contract announcements no longer compensate automatically for losses and concentration.

What the Nscale IPO Numbers Do Not Settle

Nscale’s filing confirms extraordinary demand, but it leaves critical questions about margins, financing, governance, and customer durability unanswered.

The first uncertainty concerns profitability at scale. Revenue growth alone does not show whether Nscale’s contracts will produce enough cash to cover financing, depreciation, power, operations, and continued expansion.

Large customers possess significant negotiating leverage. Microsoft and Anthropic can offer volume and contract duration while demanding strong performance commitments and competitive rates.

Nscale might generate substantial revenue without earning attractive margins if financing and operating costs consume most of the payments. Investors need project-level economics to evaluate that possibility.

The second uncertainty involves financing the Anthropic agreements. Nscale says it is pursuing multiple sources, but no binding commitments were in place when the registration statement was filed.

This does not mean the company will fail to finance the project. Nscale has repeatedly raised equity and debt. It means the contract’s value depends on a financing process that remains incomplete.

The third uncertainty is customer diversification. Nscale has agreements involving other companies, including OpenAI and robotics developer Figure, alongside projects in several markets.

Those relationships broaden the company’s network. They remain much smaller than the Microsoft and Anthropic commitments within the disclosed contract book.

Diversification cannot be measured by the number of logos alone. Investors need to know how much revenue each customer contributes, how long agreements run, and whether contracts share common dependencies.

Several projects may depend on the same chip supplier, financing market, or power schedule. A larger customer list therefore does not always create the operational diversity its headline suggests.

The fourth uncertainty involves internal controls. A company scaling this quickly must expand finance, compliance, information technology, and reporting systems alongside its physical infrastructure.

Nscale disclosed material weaknesses concerning entity-level and business-process controls, information technology controls, and finance staffing. It plans remediation work across 2026 and 2027.

Such disclosures are not unusual among companies approaching an IPO. They still matter because Nscale manages numerous subsidiaries, financing arrangements, construction projects, and long-term customer agreements.

The fifth issue is valuation. Reports have linked the offering to a potential valuation around $35 billion and fundraising targets between approximately $2 billion and $3 billion.

The final terms were not established in the initial filing. Those numbers should therefore be treated as expectations rather than completed transaction details.

Nscale was valued at $14.6 billion when it announced its Series C round earlier in 2026. A much higher IPO valuation would require investors to credit recent contracts and future delivery heavily.

The company’s loss complicates conventional valuation methods. Earnings multiples provide little help, while present revenue remains small relative to the proposed value.

Investors will probably focus on backlog, expected revenue growth, financed megawatts, and projected cash flow. Each measure depends on assumptions about timing and completion.

A high valuation could still prove justified if Nscale activates capacity quickly and retains its anchor customers. It becomes harder to defend if deployments slip or financing consumes too much of the economics.

The concentration analysis frames the listing as another test of Wall Street’s appetite for tightly connected AI bets. That is accurate, but concentration is only part of the decision.

The deeper issue is whether Nscale can turn interdependence into durable cash flow. Microsoft, Anthropic, Nvidia, lenders, and Nscale all benefit if the campuses launch successfully.

The same relationships offer limited insulation when assumptions fail. A chip delay can disrupt construction, customer acceptance, financing draws, and revenue recognition in sequence.

Investors should avoid two opposite conclusions. The contracts are not meaningless because they contain conditions, and they are not guaranteed revenue because they carry large stated values.

They are commercial commitments attached to an immense delivery program. The Nscale IPO asks public shareholders to finance that program while accepting its concentrated exposure.

Three Signals Will Decide Whether the Bet Works

The strongest evidence for Nscale will come from financed projects becoming operating infrastructure and producing recognized revenue.

The first signal is financing for the Anthropic buildout. Nscale must obtain qualifying capital for the Monarch Compute Campus infrastructure and its required GPU equipment.

A binding financing package on workable terms would strengthen the company’s central claim. It would show that lenders and capital partners believe the contracts can support the project.

A delay, expensive financing, or a reduced project would weaken the case. It would demonstrate that customer demand alone cannot overcome capital-market limits.

The second signal is the conversion of Microsoft commitments into delivered capacity and reported revenue. Investors should compare new revenue against deployment milestones rather than the full theoretical contract total.

Progress would appear through energized facilities, installed systems, customer acceptance, and higher service revenue. It should also reduce the gap between Nscale’s operating scale and its backlog.

Repeated schedule changes would carry more weight than another contract announcement. The company already has demand on paper. Its scarce resource is dependable execution.

The third signal is whether customer concentration declines as the business expands. Nscale does not need Microsoft or Anthropic to become smaller customers in absolute terms.

It needs other customers to grow quickly enough that a change by either anchor customer would no longer define the entire company. That requires material revenue diversification, not only additional partnerships.

Investors should watch future filings for the share of revenue generated by the largest customer. They should also examine how much contracted value remains conditional on common financing or delivery dependencies.

These signals will matter more than the first trading day. Strong demand for shares can validate an offering price, but it cannot validate projects scheduled across several years.

The Nscale IPO will ultimately test whether public investors accept concentrated exposure as the unavoidable cost of building AI infrastructure at exceptional speed.

For developers and enterprise buyers, the outcome also matters beyond one stock. Infrastructure funding affects available GPU capacity, regional supply, service competition, and the resilience of AI workloads.

For investors, the practical task is equally direct. Track financing, delivered capacity, and customer mix before treating the $103 billion contract figure as an operating result.

Nscale has already established that Microsoft and Anthropic want enormous amounts of compute. The remaining question is whether it can fund, build, and operate that capacity without allowing two customers to determine every outcome.